Ancient Monotheism

"It may be declared that, in this strict sense, no savage tribe of monotheists has been ever known." — E. B. Tylor, 1871¹

"From all this evidence it does not appear how non-polytheistic, non-monarchical, non-Manes-worshipping savages evolved the idea of a relatively supreme, moral, and benevolent Creator, unborn, undying, omniscient, and omnipresent." — Andrew Lang, 1898²

"In this being, although supernatural, there is no trace of a divine nature… But in saying this I must guard myself from being thought to imply any primitive revelation of a monotheistic character." — A. W. Howitt, 1904, on the Australian All-Father — writing as Lang's own principal source³

Summary

The claim under test is a piece of folk knowledge that almost everyone absorbs without ever seeing it argued:  religion climbed a ladder.  Animism first, then spirits, then gods, then one god.  Monotheism sits at the top, so it must be late — which is why an ancient text that talks about a single creator is usually assumed to be younger than it claims.

That ladder is dead, and creationists did not kill it.  Anthropology abandoned unilinear cultural evolution generally, and the one modern attempt to reconstruct forager religion with phylogenetic methods found that belief in a high god does not correlate with any other religious trait — not with ancestor worship, not even with belief in an afterlife.  Whatever else it is, it is not the last rung of a sequence.⁴

But the apologetic counter-claim — Andrew Lang's and Wilhelm Schmidt's "original monotheism," revived in 2013 by Winfried Corduan and put to a wide audience by Michael Jones in 2020 — does not survive contact with the data either.⁵  The Ethnographic Atlas holds 1,291 societies, 775 of them coded for the high-god variable that this whole literature runs on.  Recounted here from the D-PLACE release:

Foragers are the least likely people on earth to hold a high god, not the most.  Lang and Schmidt said the opposite, and they were wrong about it.

Six findings, and they cut in different directions.

1. The gradient runs the wrong way for Schmidt.  Belief in a high god rises steadily with political complexity.  Across 751 societies the rank correlation between jurisdictional levels above the local community and the high-god variable is rho = 0.44.  Among societies with no political authority beyond the village, 56% have no high god at all;  among societies with four levels of hierarchy, 66% have an active, morally policing one.  Durkheim's version of this — "the height of the god depends on the breadth of the polity," in Hiatt's phrase — is what the numbers say.⁶

2. But the kind of high god is what varies, and that rescues part of the old case.  Split it differently:  given that a society has a high god at all, what sort is it?  Among foragers, 75.3% of those gods are otiose — a creator who made the world and withdrew.  Among societies with four levels of political hierarchy, only 18.2% are, and 81.8% are active moral enforcers.  The withdrawn creator is not built by states.  What states build is the god who watches you.

3. A large part of the "moralizing high god" signal is just Islam and Christianity.  Of the 198 societies coded as having an active, morality-supporting high god, 125 — 63.1% — sit in the Islamic and Christian belt running from Morocco to Central Asia.  The pastoralist cell that scores 72.6% is Bedouin, Berber, Somali, Tuareg and Beja, plus Icelanders and Greenlanders.  Restrict the analysis to the Americas, Australia, Papua and the Pacific, where neither religion had arrived, and the complexity gradient halves (rho falls from 0.44 to 0.22).  It does not vanish — but a literature that treats this variable as a measurement of independent cultural evolution is partly measuring missionary reach.

4. The variable cannot see the thing Lang was arguing about.  Ethnographic Atlas code 1 is not "absent."  It is "a high god absent or not reported in substantial descriptions of religious beliefs."⁷  Lang's entire case was that the All-Father was initiation lore, deliberately withheld from women, children and outsiders, and that ethnographers who asked directly got nothing.  A code that merges "these people have no creator god" with "nobody wrote one down" is structurally blind to his hypothesis.

5. And the Atlas has an Australia-shaped hole exactly where the argument was fought.  Thirteen Australian societies are in the Ethnographic Atlas;  ten are coded for high gods;  nine of those ten are coded "absent," and the tenth (Tiwi) "otiose."  But look at which societies:  Northern Aranda, Walbiri, Gidjingali, Yir Yoront, Anindilyagwa, Yolngu, Tiwi, Diyari, Murriny Patha, north-western Tasmanians.  Not one south-eastern group.  The Kurnai, Kulin, Wiradjuri, Kamilaroi and Yuin — the peoples whose Baiame, Bunjil, Daramulun and Mungan-ngaua were the entire evidential basis of Lang's and Schmidt's case — are absent from the sample, because by the time systematic ethnography existed those societies had been destroyed.  Neither side of this argument usually notices that the modern database excludes the classic evidence.

6. Documented monotheism in the literate world is far older than 500 BC, and this is where the strongest version of the argument lives.  Erik Hornung, who spent a career arguing that Egyptian religion was not monotheistic, says of Akhenaten's reform in the 1350s BC:  "Now, for the first time in history, the divine has become one, without a complementary multiplicity."⁸  W. G. Lambert, reading a Late Babylonian god-list that makes Ninurta, Nergal, Nabu, Sin, Shamash and Adad into aspects of Marduk, concludes that "there is no escape from the conclusion that this is an assertion of monotheism," while noting that no such document can be proved earlier than 800 BC.⁹  Xenophanes of Colophon, around 530 BC, wrote "One god, greatest among gods and men, in no way similar to mortals either in body or in thought."¹⁰  These are not fringe readings;  they are the standard handbooks.

The deepest problem is that both sides settled the question by definition, and neither noticed.  Tylor conceded outright that peoples "of North and South America, of Africa, of Polynesia" acknowledge "a Supreme Creator" — and then tightened the definition of monotheism until they did not count.¹  Schmidt did the mirror image:  where lesser beings are created by the Supreme Being and subordinate to him, he ruled that "such superior beings do not deserve the title of gods," and so the religion "must be declared… still completely monotheistic."¹¹  The word itself is a seventeenth-century English coinage, minted by the Cambridge Platonist Henry More in 1660 in the middle of an argument about pantheism.¹²  Applying it to a Sumerian or an Aranda is an act of translation, not of measurement.  The Encyclopaedia Iranica says it plainly about Zoroaster:  "depending on what one means all three labels" — monotheist, dualist, polytheist — "can be made to apply."¹³

Verdict.  Original monotheism cannot be established and should not be claimed.  The oldest reconstructible trait of forager religion is animism, not a creator;  Schmidt's method is dead;  and the simplest societies on record are the ones least likely to name a high god.  Anyone arguing that mankind began monotheistic is arguing past the evidence.

But the claim the argument was invented to answer is dead too, and that matters more than it looks.  There is no law of religious evolution that makes an early monotheistic text anachronistic.  Belief in one supreme creator turns up in about two of every five foraging societies, on every inhabited continent, uncorrelated with everything else in their religion;  it is documented in Egypt eight centuries before the Babylonian exile and in Babylon three centuries before it;  and a Greek was arguing for it philosophically while the Judean monarchy still stood.  Whether Israel's monotheism is early or late has to be decided from Israel's own evidence.  It cannot be decided by a stage theory, because there is no stage theory left.

Every source is catalogued with what it says and which way it cuts in a companion page:  the literature.

The claim, stated fairly

Two arguments run in opposite directions, and both are older than they look.

The skeptical argument.  Religion has a natural history.  It begins with the attribution of soul and intention to everything — animism.  Souls of the dead become ghosts, ghosts become gods, one god rises above the rest, and eventually the rest are denied.  Monotheism is therefore a late and sophisticated product, and a text that shows it is late, or has been edited late.  Applied to the Bible, this yields the standard reconstruction:  the patriarchs' religion was at most henotheistic, Israel's Yahweh began as one god among the sons of El, and unambiguous monotheism — "besides me there is no god" — first appears in Isaiah 40–55, during or after the Babylonian exile.

The apologetic counter-argument.  In its strong form:  mankind began knowing one God, and polytheism is a decay product.  Andrew Lang put this forward in 1898 against his own former teacher's theory;  Wilhelm Schmidt spent forty years and twelve volumes documenting it;  Winfried Corduan revived it in 2013;  and Michael Jones ("InspiringPhilosophy") gave it its widest modern hearing in a 55-minute documentary in March 2020.⁵  In its weak form, which is the one Jones actually argues for and states explicitly at the end of his video, the claim is only this:  monotheistic ideas are attested in very simple societies and in very ancient texts, so an early monotheistic text is not automatically anachronistic.

This article tests both.  It goes badly for the strong form and well for the weak one.

Tylor's ladder, and what Tylor actually said

Edward Burnett Tylor published Primitive Culture in 1871 and became the founding figure of British anthropology.  His method was to define religion as broadly as possible so that no people fell outside it:  "It seems best to fall back at once on this essential source, and simply to claim, as a minimum definition of Religion, the belief in Spiritual Beings."¹  He named that minimum animism.

The evolutionary sequence that everyone remembers is Tylor's, and it is real.  But his treatment of high gods is far more careful than the summaries suggest, and reading him is the fastest way to see how the whole argument went wrong.

Tylor did not deny that "savages" have supreme creator gods.  He says the opposite, in as many words:

Races of North and South America, of Africa, of Polynesia, recognizing a number of great deities, are usually and reasonably considered polytheists, yet under this definition their acknowledgement of a Supreme Creator, of which various cases will here be shown, would entitle them at the same time to the name of monotheists.¹

What he then does is tighten the definition until they stop qualifying:

To mark off the doctrines of the lower races, closer definition is required, assigning the distinctive attributes of deity to none save the Almighty Creator.  It may be declared that, in this strict sense, no savage tribe of monotheists has been ever known.  Nor are any fair representatives of the lower culture in a strict sense pantheists.  The doctrine which they do widely hold, and which opens to them a course tending in one or other of these directions, is polytheism culminating in the rule of one supreme divinity.¹

That is the whole dispute in one paragraph.  The datum — a supreme creator, widely acknowledged, in simple societies — is not contested by Tylor.  What is contested is the label.

He also supplies, in advance, the objection that would be used against Lang for the next century:

Among low tribes who have been in contact with Christianity or Mohammedanism, how are we to tell to what extent, under this foreign influence, dim, uncouth ideas of divine supremacy may have been developed into more cultured forms, or wholly foreign ideas implanted?  We know how the Jesuit missionaries caught and trained into their own theology the native Canadian thought of a Great Manitu…¹

And he gives his mechanism for how a supreme deity arises — three routes, all of them from below:

…the simple process of raising to divine primacy one of the gods of polytheism itself.  Even the system of the manes-worshipper has been stretched to reach the limit of supreme deity, in the person of the primæval ancestor.  More frequently, it is the nature-worshipper's principle which has prevailed, giving to one of the great nature-deities the precedence of the rest.¹

Tylor even knew the Vedic line that gets quoted at him — he prints Rig Veda 1.164.46 in the original and in Max Müller's translation, "That which is One the wise call it in divers manners."¹

So the honest summary of Tylor is this:  he accepted the ethnographic data, redefined the terms so that the data could not count as monotheism, and proposed three developmental routes to explain the data away.  Section 8 below tests those routes against a database he never saw.  Two of them fail.

Lang's revolt, and why it went nowhere

Andrew Lang was a Tylorian.  In The Making of Religion (1898) he opens by stating his own former position with unusual clarity:

Man derived the conception of 'spirit' or 'soul' from his reflections on the phenomena of sleep, dreams, death, shadow… Worshipping first the departed souls of his kindred, man later extended the doctrine of spiritual beings in many directions.  Ghosts, or other spiritual existences fashioned on the same lines, prospered till they became gods.  Finally, as the result of a variety of processes, one of these gods became supreme, and, at last, was regarded as the one only God.²

He then breaks with it, on one specific ground:  the supreme beings reported among the simplest peoples do not look like promoted ghosts.  They are not fed, they do not need sacrifice, they are often not even worshipped, and they are moral.  His conclusion, at the end of the survey:

From all this evidence it does not appear how non-polytheistic, non-monarchical, non-Manes-worshipping savages evolved the idea of a relatively supreme, moral, and benevolent Creator, unborn, undying, omniscient, and omnipresent.²

Lang was scrupulous about the contamination problem — more so than his critics allowed.  He opens his chapter on high gods by warning against exactly the bias he was accused of:

The missionary who does not see in every alien god a devil is apt to welcome traces of an original supernatural revelation, darkened by all peoples but the Jews.  We shall not, however, rely much on missionary evidence, and, when we do, we must now be equally on our guard against the anthropological bias in the missionary himself.²

And his account of how monotheism decays into polytheism is the best thing in the book, because it is a mechanism rather than a lament:

A moral creator in need of no gifts, and opposed to lust and mischief, will not help a man with love-spells, or with malevolent 'sendings' of disease by witchcraft;  will not favour one man above his neighbour, or one tribe above its rivals, as a reward for sacrifice which he does not accept, or as constrained by charms which do not touch his omnipotence.  Ghosts and ghost-gods, on the other hand, in need of food and blood, afraid of spells and binding charms, are a corrupt, but, to man, a useful constituency.²

That is a real hypothesis with a testable shape:  a god who cannot be bargained with will be neglected in favour of gods who can.  Section 8 finds something adjacent to it in the data — but not quite it.

Why nobody listened.  Two reasons, and only one of them is prejudice.

The first is that Lang's argument was ignored because of the company it kept.  The Making of Religion has seventeen chapters.  Seven of them — "Science and 'Miracles'," "Opening the Gates of Distance," "Crystal Visions, Savage and Civilised," "Anthropology and Hallucinations," "Demoniacal Possession," "Fetishism and Spiritualism," and half of "Anthropology and Religion" — argue that telepathy, clairvoyance and veridical hallucination are real, with appendices on poltergeists and crystal-gazing.  Lang saw the problem and said so on his second page:  the two halves of his book "do not depend on each other," and the first "may be regarded as fantastic, or improbable, or may be 'masked' and left on one side."²  Readers took him at his word and masked the whole book.  A serious ethnographic argument was published inside a defence of psychical research, and paid for it.

The second reason is that Lang's own best source disagreed with him about what the data meant.

Howitt: what Lang's principal source actually concluded

A. W. Howitt spent decades among the Aboriginal peoples of south-eastern Australia and published The Native Tribes of South-East Australia in 1904.  He is the man Lang leaned on hardest, and he is the man modern popular treatments quote.

On the question of whether the All-Father belief was borrowed from missionaries, Howitt is emphatically on Lang's side, and his reasoning is worth having in full because it is the strongest form of the argument.  E. M. Curr had claimed that writers reporting Aboriginal knowledge of God "have been imposed upon," and that the blacks, "having heard the missionaries, were not slow to invent what he might call kindred statements with aboriginal accessories, with a view to please and surprise the whites."³  Howitt's reply:

That the blacks had no knowledge of God, and that they did not practise prayer, is quite true;  but they had, and have beliefs such as those which I have recounted, which were evidently unknown to Mr. Curr, and which he was not likely to learn from the manner of his inquiries "concerning God and the next world."  Nor would those he questioned voluntarily tell him, even if they understood his meaning, the sacred and secret teachings of the initiation ceremonies.  In my experience, men have in such circumstances put on an appearance of dense stupidity, or have resorted to absolute denial of any knowledge, and even to lies.³

He adds a control that is genuinely good evidence.  Missions had operated among the Narrang-ga, Parnkalla, Dieri and Arunta for long periods, preaching in the native language.  If the All-Father were a missionary import, those are the tribes that should have it.  They do not.  "How is it, if we are to assume that the All-father belief in the south-east has been due to missionary teachings, that there has not been a similar adoption of it by the western tribes?"³

And then Howitt rejects the theological reading outright.  This is the part that gets left out:

In this being, although supernatural, there is no trace of a divine nature.  All that can be said of him is that he is imagined as the ideal of those qualities which are, according to their standard, virtues worthy of being imitated.  Such would be a man who is skilful in the use of weapons of offence and defence, all-powerful in magic, but generous and liberal to his people…³

He then proposes his own developmental sequence — ghost, then a sky-tribe of ancestral ghosts with a Headman, then the All-Father as that Headman writ large — and closes the door explicitly:

Such seems to me the probable course of development of this belief, which moreover I am satisfied has been locally evolved, and not introduced from without.  But in saying this I must guard myself from being thought to imply any primitive revelation of a monotheistic character.  What I see is merely the action of elementary thought reaching conclusions such as all savages are capable of, and which may have been at the root of monotheistic beliefs.³

Howitt does hand Lang one substantial point against Tylor and Spencer:  "Australian evidence seems to carry us back to a stage before ancestors came to be worshipped."³  Ancestor worship is not the root.  That specific piece of Tylor's ladder was already broken in 1904, by a hostile witness, and it is broken again by the phylogenetics in section 9.

Anyone using Howitt to argue for original monotheism is using half of him.  The right way to cite him is:  the belief is indigenous, the belief is not revelation, and the belief is not what Genesis is talking about.

Schmidt's twelve volumes, and the method that sank them

Wilhelm Schmidt (1868–1954) was a Catholic priest of the Society of the Divine Word, a first-rate linguist, and the most industrious defender the original-monotheism thesis has ever had.  Der Ursprung der Gottesidee ran to twelve volumes and over 11,000 pages.  The English Origin and Growth of Religion (1931;  2nd ed. 1935) is the accessible summary of his position.

Schmidt's claim is stronger than Lang's.  Not "supreme beings exist among simple peoples," but "the religion of the primitive culture is a monotheism":

That the Supreme Being of the primitive culture is really the god of a monotheism, and that the religion which includes him is genuinely monotheistic — this is the position which is most attacked by a number of authors.  To this attack we may reply that there is a sufficient number of tribes among whom the really monotheistic character of their Supreme Being is clear even to a cursory examination.¹¹

And here is how he handles the obvious objection, that these peoples also have plenty of other supernatural beings:

If they are more or less clearly stated to have been created by him, and consequently have acquired their attributes and powers from him;  or, still more, if their occupations and positions are assigned to them by him;  but most of all if the Supreme Being still oversees and regulates the exercise of these functions, then we must declare such a religion to be still completely monotheistic.  Such superior beings do not deserve the title of gods.¹¹

That is Tylor's move performed in reverse.  Tylor narrowed "monotheism" until nothing counted;  Schmidt widened it until a great deal did.  Neither move is a discovery about the world.

Three things sank Schmidt, and only one of them is prejudice.

The method.  Schmidt's conclusions depended on the culture-historical method — the Kulturkreislehre, which sorted the world's peoples into "culture circles" and ranked them by antiquity.  Corduan, its most determined modern defender, concedes that it "is often completely left out of summaries of his work, possibly because for various reasons it has remained obscure."¹⁴  The problem is not obscurity.  The problem is that the ranking of cultures by primitiveness was doing the load-bearing work, and it was constructed partly from the traits it was then used to explain.  Modern anthropology has abandoned the framework, and Michael Jones says so in his video rather than hiding it.⁵

The circularity of "primitive."  Schmidt named the Pygmies of Central Africa among the most primitive peoples alive, and built on their monotheism.  Later fieldwork found many of those groups polytheistic.  Jones concedes this too, in his own video description.  When the reference class can be revised after the fact, the result is not a measurement.

The teleology.  Mircea Eliade, who admired Schmidt's scholarship, put the objection best:

What cannot be accepted in Schmidt's reconstruction is his exclusively rationalistic approach.  He asserts that primitive man discovered the idea of God through a logical quest for a cause… Schmidt was inclined to think that all the irrational elements represent a "degeneration" of the genuine, primordial religion.  The truth is that we do not have any means to investigate this "primordial religion."  Our oldest documents are relatively recent.¹⁵

Eliade's own summary judgment is the most defensible sentence anyone has written on the subject, and it is the position this article ends up at as well:

It is true that the belief in High Gods seems to characterize the oldest cultures, but we also find there other religious elements.  As far as we can reconstruct the most remote past, it is safer to assume that religious life was from the very beginning rather complex, and that "elevated" ideas coexisted with "lower" forms of worship and belief.¹⁵

One correction to the common apologetic account of Schmidt's reception:  it is often said that he was ignored because he was a priest.  Eliade does speculate along those lines.¹⁵  But Eliade also records, in the same paragraph, that Robert Lowie and Paul Radin — neither of them Catholic, neither of them a degenerationist — "acknowledged the belief in the existence of supreme beings among the most archaic peoples."¹⁵  The datum was never really in dispute.  The interpretation was.

Missionary contamination: the honest state of the question

This is the objection that decides how much of the ethnographic evidence survives, and it is usually argued badly by both sides.

What the borrowing hypothesis has going for it.  It has Tylor's Great Manitou and Tupan examples, both documented.¹  It has the real fact that most of the reports come from a period of intense missionary activity.  And, in the Australian case specifically, it had the whole weight of professional opinion:  Tylor devoted his 1891 presidential address to the Anthropological Institute, "On the Limits of Savage Religion," to the argument that reports of native monotheism "all… turn out to be consequences of ideas implanted in the minds of the natives by missionaries."⁶

What it has going against it.  Lang's four points, as summarised by L. R. Hiatt, the leading modern historian of the controversy:

First, Aboriginal High Gods have been recorded in circumstances strongly suggesting that the beliefs were in existence prior to contact with Europeans… In order to maintain that Aboriginal notions of deity are the product of proselytization, one would have to assume that Christian beliefs caught on at their points of introduction and (like smallpox) travelled ahead of the invaders.  Second, detailed information about Aboriginal High Gods has consistently come from initiated adult males, on the understanding that it should not be divulged to women and children.  If it were true that the deities of the male cults were borrowed from the missionaries, how would we reconcile the men's desire for secrecy with the fact that knowledge of Christian doctrine is freely available to all?  Furthermore, male elders in Aboriginal society are notoriously conservative.  Why, especially in the matter of religion, should they be so ready to abandon old beliefs and embrace the new?⁶

Add Howitt's control — missions worked hardest among the western tribes, which lack the belief — and the strong borrowing hypothesis is in serious trouble.³  Corduan's summary is blunt and, on this narrow point, fair:

The notion that the belief in a single supreme being spread lightning fast, ahead of the missionaries, can only be characterized as Tylor's bad story made worse.  If these supreme beings were inconsistent with the rest of these cultures, if in some tribes they never received any worship, and if they were not associated with any benefits from believing in them, it makes no sense that all of these tribes picked them up instantaneously.  It is utterly implausible, which may just be the reason there is no evidence for it.¹⁴

But there is a second objection that survives the first, and apologetic treatments almost never mention it.  In 1898 Sidney Hartland granted, for argument, that the beliefs were indigenous — and then asked what they actually contained.  Hiatt's summary:

Indeed, once we begin to look carefully at the myths and legends, comparison with civilized conceptions of divinity becomes embarrassing.  Daramulan is a cannibal whose name means 'leg-on-one-side' or 'lame'.  Bunjil, with his two wives and six sons, was blown into the sky by wind released from bags by an angry man who later became a bird.  Baiame tripped and fell on his face one day while pursuing an emu.  In mystery cults he is shown carved in the earth in this humiliating posture, together with a print of his hand left in the ground when he tried to break the fall.  Such childish tales bear no comparison with the sublime conception of the Creator as set forth in the book of Genesis.⁶

Lang's reply was that a lofty conception and a ridiculous mythology routinely coexist — Greece being the obvious case — and that the sacred rites, not the campfire stories, carry the theology.⁶  That reply is reasonable.  It is not a refutation.  It leaves the honest position roughly where Hartland left it:  the beliefs are native, and they are not the God of Genesis.

How the controversy actually ended:  it didn't.  It was abandoned.  Baldwin Spencer and Francis Gillen's fieldwork among the Aranda of Central Australia supplied what looked like decisive evidence for the other side — Spencer wrote to Frazer from the Gulf of Carpentaria that "we cannot find a trace of any belief right through the Central tribes from Port Augusta in the south to the Gulf in the north in a being who could be called a deity."⁶  Then, in 1927, Spencer revised the book after another visit to Alice Springs and quietly added a section on Numbakulla, whom the Aranda "regarded as 'the supreme ancestor, overshadowing all others'," who created the mountains and rivers and the sacred icons, and who "rose into the sky and disappeared forever."  Hiatt's comment is the epitaph:  "In providing scholars with this new information, Spencer felt no need to correct any earlier impressions he may have given.  By 1927 the High God controversy, like most who had participated in it, was dead."⁶

Nor did the Australian material settle down afterwards.  When anthropologists returned to the question in the mid-twentieth century the candidate for a pan-continental high god had shifted to the Rainbow Serpent — and in 1978 Kenneth Maddock argued that a single, well-defined continental Rainbow Serpent is itself "an anthropological fiction," since the accumulated evidence shows no constancy of form, powers or role, and in many groups the most important deities have nothing to do with the rainbow at all.⁶  Both sides of the original argument had been generalising across a continent from a handful of regions.

Two smaller points, for the record.  Michael Jones renders Lang's degeneration passage as "a moral creator in no need of gifts";  Lang wrote "a moral creator in need of no gifts."²  And the sentence "Baiame thus closely resembles one aspect of the god of the Hebrews," attributed in the video to Radcliffe-Brown, was not verified here against Radcliffe-Brown's own text and is not used in this article.

The recount: what 1,291 societies actually show

Everything quantitative in this field — Swanson 1960, Norenzayan's Big Gods, Botero 2014, Watts 2015, Peoples 2016 — runs on one variable:  Ethnographic Atlas variable 34, "Religion:  high gods," coded by George Peter Murdock and successors for 1,291 societies.  The variable was recomputed here from the D-PLACE CLDF release.  The script is published beside this article and runs in under a second.¹⁶

What the variable is, and what it cannot see

Four states, in Murdock's own wording:

1. Absent — "A high god absent or not reported in substantial descriptions of religious beliefs"
2. Otiose — "A high god present but otiose or not concerned with human affairs"
3. Active, not moral — "A high god present and active in human affairs but not offering positive support to human morality"
4. Active, moral — "A high god present, active, and specifically supportive of human morality"⁷

Read code 1 again.  It merges two very different findings:  this people has no creator god, and nobody recorded one.  For most purposes that conflation is harmless.  For this question it is fatal, because the central claim on the other side is that the high god was esoteric knowledge that ethnographers did not get.  Howitt's informants, on his own account, met direct questions with "an appearance of dense stupidity, or… absolute denial of any knowledge, and even to lies."³  A variable that scores that as "absent" cannot test the hypothesis.  Every number below inherits that bias, and every number below therefore understates high gods by an unknown amount.  It is stated once here and not repeated.

775 of the 1,291 societies are coded.  The 516 blanks are not random either — they are the societies whose ethnography was thinnest.

The headline numbers

Nearly two-thirds of surveyed human societies acknowledge a supreme creator.  That is the datum Tylor conceded and then defined away, and it is still there.

Now split it by subsistence:

Rank correlation between subsistence type and the high-god variable:  rho = 0.251 over 775 societies.  Lang and Schmidt predicted a negative correlation.  They got a positive one.

The prediction fails hardest where it should have succeeded best.  Take the very simplest societies in the Atlas — nomadic or seminomadic foragers with no political authority above the local community, the closest thing to Lang's test case that exists.  There are 94 of them.  62.8% have no high god recorded, and 3.2% have an active moral one.

The complexity gradient

The strongest correlate is not subsistence but politics:

rho = 0.438 across 751 societies.  Mean community size gives rho = 0.377;  class stratification rho = 0.275;  settlement pattern only rho = 0.098.

This is Durkheim's result, obtained a century later on a database he never saw.  Hiatt's paraphrase of Durkheim's 1912 argument — that Australian high gods are gods of inter-tribal initiation cults, and that "the height of the god depends on the breadth of the polity" — is a compact statement of what the numbers show.⁶

A caution about those p-values.  Societies are not independent observations:  neighbours share ancestry and borrow from each other.  This is Galton's problem, it is a hundred and thirty years old, and it means the conventional significance tests here are far too generous.  The effect sizes are what matter.  The within-region breakdown printed by the script is a crude control, and the Abrahamic-belt test below is a better one.

Which kind of high god — the finding that rescues part of the old case

Everything above asks whether a society has a high god.  Ask instead:  given that it has one, what kind is it?

Where a foraging society has a creator, three times in four he made the world and withdrew.  Where a large state has one, four times in five he is watching and enforcing.  The gradient is monotonic across all five levels of political hierarchy and it is steep — 71% otiose at the bottom, 18% at the top.

This is the deus otiosus, and it is the single most interesting thing in the data.  It is not what Lang and Schmidt claimed — they wanted a moral, active creator at the bottom, and the moral, active creator is at the top.  But it is not what Tylor's mechanism predicts either.  Tylor's three routes were:  promote a polytheistic god, deify the first ancestor, or elevate a nature-deity.¹  All three produce a god who is doing something — ruling, receiving sacrifice, controlling weather.  None of them naturally produces a creator whose defining feature is that he has stopped intervening and receives no cult.  A withdrawn creator is a poor promotion candidate and a useless ancestor.

Lang saw this and made it his central point:  the high gods of the simplest peoples are the ones who need no gifts and grant no favours, and that is exactly why they get neglected.²  The data do not confirm his history.  They do confirm his typology.

The Abrahamic confound

Here is a check that the "Big Gods" literature does not usually run.  Of the 198 societies coded "active, supporting morality," how many are simply Muslim or Christian?

Counting the thirteen Ethnographic Atlas regions that Islam and Christianity had reached by the ethnographic present — Northern Africa, Northeast Tropical Africa, Western Asia, the Arabian Peninsula, Middle Asia, the Caucasus, Eastern, Middle, Northern, Southeastern and Southwestern Europe, the Western Indian Ocean and the Indian Subcontinent — the answer is 125 of 198, or 63.1%.  Drop the Indian Subcontinent, which is the arguable inclusion, and it is still 119, or 60.1%.

Look at the pastoralist cell that scores 72.6% active-moral.  Its members are the Rwala, Mutair, Kababish, Shuwa, Hadendowa, Bisharin, Beni-Amer, Tigre, Afar, Somali, Tuareg Ahaggar, Chaambra, Regeibat, Wodaabe Fulani, Kazakh, Bakhtiari, Basseri — plus Icelanders, Sami, Ossetians, Abkhaz and West Greenlanders.  That is not a discovery about pastoralism.  That is a map of Islamic and Christian mission history.

So re-run the gradient on the twenty-five Ethnographic Atlas regions the Abrahamic religions had not reached at the ethnographic present — the Americas, Australia, Papua, Malesia and the Pacific.  334 societies:

The gradient survives, at about half strength.  So social complexity really does predict moralizing gods independently of Islam and Christianity — Botero's and Watts's finding is not an artefact.  But roughly half of the raw effect, and nearly two-thirds of the raw count, is the spread of two proselytizing religions with armies and empires behind them.  Any sentence of the form "moralizing gods emerged as societies grew complex" is describing that mixture, and should say so.

The Australia-shaped hole

The Ethnographic Atlas contains thirteen Australian societies.  Ten are coded for high gods.  Nine of the ten are coded absent;  the tenth, the Tiwi, is coded otiose.  On its face that is a rout for Lang.

Then look at which ten:  Northern Aranda, Walbiri, Gidjingali, Yir Yoront, Anindilyagwa, Yolngu (Dhuwal), Tiwi, Diyari, Murriny Patha, north-western Tasmanians.  Central desert, Arnhem Land, Cape York, the Kimberley, Tasmania.

Not one south-eastern society is in the sample.  Not the Kurnai, not the Kulin, not the Wiradjuri, not the Kamilaroi, not the Yuin, not the Ngarrindjeri.  Those are the peoples of Mungan-ngaua, Bunjil, Baiame and Daramulun.  They are the entire evidential basis on which Lang built his case and Schmidt built his system, and they are missing from the modern database — because by the time anthropology had standardised its methods, those societies had been destroyed by settlement.  The one south-eastern group in the Atlas, the Ngiyambaa, has no high-god code at all.

This does not vindicate Lang.  Howitt himself denied that the All-Father was a god in the theological sense, and the reports come from a small and contested corpus.  But it does mean that the quantitative literature and the historical controversy are, in the Australian case, not talking about the same societies — and neither side seems to have said so.  The number to keep from Australia is not "9 of 10 absent."  It is "0 of 13 sampled from where the argument happened."

What the phylogenetics says

In 2016 Hervey Peoples, Pavel Duda and Frank Marlowe did the obvious modern thing:  took 33 hunter-gatherer societies, built a time-calibrated supertree from published phylogenies and linguistic classification, coded seven religious traits, and reconstructed the ancestral states.⁴  It is the only serious attempt at this and it is open access.  Their traits were animism, belief in an afterlife, shamanism, ancestor worship, active ancestor worship, high gods, and active high gods — and their high-god variable is Ethnographic Atlas variable 34, the same one recounted above.

Their results go against both camps, in different places.

For Tylor, against Schmidt:  the oldest reconstructible trait is animism.

Results indicate that the oldest trait of religion, present in the most recent common ancestor of present-day hunter-gatherers, was animism, in agreement with long-standing beliefs about the fundamental role of this trait.  Belief in an afterlife emerged, followed by shamanism and ancestor worship.  Ancestor spirits or high gods who are active in human affairs were absent in early humans…⁴

That is a direct empirical answer to Corduan's 2013 claim that the original-monotheism hypothesis "stands so far unfalsified" and that "its most formidable competing hypotheses are — well, absent."¹⁴  Three years later a competing hypothesis was tested and supported.  The claim needs updating.

Against Tylor, and this is the more surprising half:  high gods do not evolve out of anything.

There is no support for coevolution of any pair of characters that includes high gods and active high gods, with the obvious exception of the two traits themselves… Surprisingly, not even belief in an afterlife shows any correlation with high gods.⁴

And in the conclusion:

However, neither high gods nor active high gods exhibit correlated evolution with the rest of the religious traits, including ancestor worship, despite Spencer's and Tylor's suggestions.⁴

Tylor's ladder said the supreme deity is the top rung — reached by promoting a ghost, an ancestor or a nature-god.  On this analysis it is not a rung at all.  High gods have the worst phylogenetic fit of any trait in the study (consistency index 0.1), meaning the trait is gained and lost rapidly and borrowed easily.  Peoples and colleagues call it "a rather 'stand-alone' phenomenon."⁴

On the specific question this article is about, their answer is: undecided.  Not "absent."

Only 39% of sample societies have the trait "high gods," and even fewer (15%) have active high gods.  The equivocal results based on maximum likelihood and absence according to maximum parsimony for the ancestral state of "high gods" suggests possible presence of belief in a single creator deity among ancestral hunter-gatherers, albeit one that is not active in human affairs (proportional likelihood = 0.05).⁴

Maximum parsimony says absent.  Maximum likelihood cannot decide.  What the two methods jointly rule out is an active creator in the ancestral state.  A withdrawn one is left open — which is, once again, the deus otiosus.

Two things to keep in mind before this paper is asked to carry too much.  The sample is 33 societies, which is small for ancestral-state reconstruction.  And it reconstructs the last common ancestor of present-day hunter-gatherers, which is not the same thing as the religion of early humanity;  every foraging society in the sample has had as long a history as everyone else.  Their 39% figure, incidentally, matches the 39.5% found above across all 185 foraging societies in the full Atlas — the two counts are independent in sample though not in source, and they agree.

The Big Gods literature, and the paper that had to be withdrawn

Since about 2013 the dominant framework has been Ara Norenzayan's:  moralizing, punishing gods are a social technology that lets strangers cooperate, and so they should appear as societies scale up.

The evidence is real but messier than the slogan.  Botero and colleagues showed in 2014 that the global distribution of moralizing high gods is predicted by ecology — resource scarcity and unpredictable environments — as well as by political complexity.¹⁷  Watts and colleagues in 2015 ran the question through 400 Austronesian cultures with phylogenetic controls and found the causal arrow pointing the other way from Norenzayan's version:  "broad supernatural punishment drives political complexity, whereas MHGs follow political complexity."¹⁸  Moralizing high gods are a consequence of scale, not a cause of it.

Then, in 2019, Nature published Whitehouse and colleagues' analysis of the Seshat historical databank under the headline "Complex societies precede moralizing gods throughout world history."¹⁹  It was widely reported as settling the matter.

It was retracted on 7 July 2021.  Beheim and colleagues showed that the result depended on how missing data had been handled.  The authors' own words:

Nevertheless, we accept that we should have labelled moralizing gods as 'absent' or 'inferred absent' rather than 'unknown' in portions of our dataset before the dates of the first appearance, rather than converting 'NAs' to zeros during the phase of analysis.  Since this Letter was published, we have thoroughly refined our data and analyses, and have found that our original conclusions are still strongly supported.  However, the differences between our revised analyses and the original Letter are substantial enough to warrant a Retraction of the original Letter.²⁰

Three of the fifteen authors — John Baines, Alan Covey and Kevin Feeney — did not agree with the retraction.²⁰  The reanalysis was published in 2022 in Religion, Brain & Behavior and reached the same conclusion by better means.²¹

Two lessons, and they are not the ones either side usually draws.  The first is that the conclusion survived:  complexity does appear to precede moralizing gods.  The second is that the most-cited empirical result in this field turned on a coding decision about absent data — the same class of decision that makes Ethnographic Atlas code 1 unusable for testing Lang.  When a field's headline finding can be flipped by how you treat blanks, treat every number in it, including the ones in this article, as provisional.

How old is documented monotheism?

Everything above is about societies without writing, reconstructed backwards.  Now the direct evidence:  texts, with dates.

Egypt: the one clear case, and it is 1350 BC

Nineteenth-century Egyptology believed Egyptian religion had begun monotheistic.  Emmanuel de Rougé, Le Page Renouf, Grébaut, Pierret and Brugsch all argued it, and Erik Hornung documents in detail how the view collapsed:  the Pyramid Texts and the pre- and early dynastic inscriptions, as they came to light, showed "especially… the multiplicity of deities," and "the idea of an original monotheism, which had for so long been the received dogma, was quietly dropped."⁸  The argument that the singular ntr, "god," in Egyptian wisdom literature reveals "a monotheism for initiates" was answered by Alfred Wiedemann in 1913 and has not recovered.⁸

Hornung's own thesis is that Egyptian religion works by complementarity:  god is "a unity in worship and revelation, and multiple in nature and manifestation," and both statements are true at once.⁸  On that reading, monotheism does not emerge from polytheism by accumulation:

…it has now become clear in logical terms why monotheism does not arise within polytheism by way of a slow accumulation of "monotheistic tendencies," but requires a complete transformation of thought patterns.  Tendencies to classify the pantheon should not be equated with an inclination toward monotheism.⁸

That is a serious blow to both sides of the popular argument — to the apologist collecting "monotheistic tendencies" in pagan texts, and to the evolutionist expecting monotheism to precipitate out of henotheism given enough time.

And then Hornung says that the transformation actually happened, once, and dates it:

In the fourteenth century B.C. there occurred the earliest attested case of this fundamental transformation of thought, in the person of Akhenaten and his teachings.⁸

Now, for the first time in history, the divine has become one, without a complementary multiplicity;  henotheism has been transformed into monotheism.⁸

He is not describing a tendency.  Akhenaten had the names of the other gods, and sometimes the plural word "gods," hacked off monuments:  "Akhenaten's aim was not just to dethrone Amun, but in principle to deny the existence of all gods except the Aten."⁸

Miriam Lichtheim's judgment, in the standard English anthology, is the same:

As early as the Old Kingdom, Egyptian religion had tended to attribute supreme power to one god, and to subordinate the other gods to him.  But while increasingly heaping attributes of universal power on the sun-god Re, the religion remained essentially polytheistic… It was Amenhotep IV who converted the supreme god into the sole god by denying the reality of all other gods.²²

The Great Hymn to the Aten, carved in the tomb of the courtier Ay at Amarna, says it in one line:

O Sole God beside whom there is none! You made the earth as you wished, you alone, All peoples, herds, and flocks²²

The honest limits.  Akhenaten's monotheism lasted less than two decades and was erased on his death — "the 'many' had overpowered the 'one'."²²  It was imposed from the throne, not grown from below.  It is not evidence of primeval monotheism;  Hornung's whole point is that it was a revolution, not a survival.  What it is is an unambiguous, dated, documented monotheism eight centuries before the Babylonian exile, produced by a culture that was polytheistic before it and polytheistic after it.

Mesopotamia: one god of whom the others are aspects

Sumerian religion had a distant creator at the head of the pantheon.  Jean Bottéro's description of An/Anu:

Thus An was the ancestor, the source, and the guarantee of power.  One could rely on his experience in times of crisis, but he left the effective exercise of sovereign authority to his son Enlil.²³

That is the deus otiosus again — the pattern found in three-quarters of the forager societies in section 8 — sitting at the top of the world's first literate pantheon.  What it is not is monotheism:  An is one god among many, and the reason he is quiet is that his son has the throne.

Far more striking is a Late Babylonian tablet, CT 24 50 (BM 47406), first published by Theophilus Pinches in 1896.  It is a two-column list.  On the left, the names of the major gods.  On the right, an aspect of Marduk:

Uraš is Marduk of planting.  Lugalidda is Marduk of the abyss.  Ninurta is Marduk of the pickaxe.  Nergal is Marduk of warfare.  Zababa is Marduk of battle.  Enlil is Marduk of lordship and consultation.  Nabû is Marduk of accounting.  Sin is Marduk who lights up the night.  Šamaš is Marduk of justice.  Adad is Marduk of rain… Marduk of everything.⁹

W. G. Lambert, the leading Assyriologist of his generation and no friend of overclaiming, is unambiguous about what this means:

…once the major members of the pantheon have become aspects of Marduk merely, there is no escape from the conclusion that this is an assertion of monotheism.  There is the one great god of whom other gods are but aspects.⁹

He supports it from three other texts:  a commentary on Marduk's Address to the Demons glossing Šamaš as "Marduk of judgement";  a hymn (KAR 337 + 304) whose four-line sections identify Adad, Sin, Šamaš and Ninurta as aspects of Marduk;  and a prayer beginning "Sin is your divinity" that works through the whole pantheon the same way.  Only one other god ever got the treatment:  a hymn to Ninurta (KAR 102, STT 118) identifies his body parts with deities — "Your eyes, lord, are Enlil and Ninlil."⁹

Lambert's verdict, with the limits he himself insists on:

Thus, the monotheistic conception of Marduk can be regarded as certain… Of course, we do not know how widespread this view may have been.  Common people worshipping the moon-god may not have been conscious of any idea that this was merely an aspect of the one god, and certainly the priests of Sin in Harran did not tolerate this notion… None of the documents preaching a monotheistic Marduk can be proved to be earlier than 800 B.C.⁹

And the field is not agreed.  J. J. Finkelstein dismissed the same tablet as a "syncretistic tendency" and "emphatically nothing to do with monotheism," and Lambert reports the general opinion after the Babel und Bibel controversy as "that the point is unproven."⁹  Bottéro, writing about Mesopotamian ideas of divine transcendence, says they remained "in an embryonic state.  It could only bear fruit in monotheism, the prerogative of Israel."²³

Two eminent Assyriologists, the same evidence, opposite verdicts.  That is not a scandal;  it is a symptom.  See section 12.

Greece: Xenophanes, c. 530 BC

Xenophanes of Colophon attacked anthropomorphic religion with an argument still in use:  "if cattle and horses or lions had hands… horses would draw the forms of the gods like horses, and cattle like cattle."  Then he stated the alternative.  Fragment 23, preserved in Clement of Alexandria:

εἷς θεός, ἔν τε θεοῖσι καὶ ἀνθρώποισι μέγιστος, οὔτι δέμας θνητοῖσιν ὁμοίιος οὐδὲ νόημα.

One god, greatest among gods and men, in no way similar to mortals either in body or in thought.¹⁰

With fragment 24 ("whole he sees, whole he thinks, whole he hears") and fragment 26 ("always he remains in the same place, moving not at all… but without toil he shakes all things by the thought of his mind").¹⁰

Kirk, Raven and Schofield are careful about the phrase "among gods":

'Greatest among gods and men' in 173 should not be taken literally;  men are mentioned by a 'polar' usage… This is simply an emphatic device, and for the same reason the plural of 'gods' need not be intended literally… It seems very doubtful whether Xenophanes would have recognized other, minor deities as being in any way related to the 'one god', except as dim human projections of it.¹⁰

Xenophanes was born about 570 BC.  His one god is philosophical rather than cultic, and he did not found a religion.  But he is a documented, dated, argued monotheism in Greek, from roughly the generation of the Babylonian exile and independent of it.

Iran: the case where the date is the whole problem

Zarathustra's Gathas address Ahura Mazdā as the creator who fixed the paths of the sun and stars and made the Good Mind.  Whether that is monotheism is disputed, and so is the date, and the two disputes are separate.

On the date, the Encyclopaedia Iranica:

Controversy over Zaraθuštra's date has been an embarrassment of long standing to Zoroastrian studies.  If anything approaching a consensus exists, it is that he lived ca. 1000 BCE give or take a century or so, though reputable scholars have proposed dates as widely apart as ca. 1750 BCE and "258 years before Alexander."¹³

That is a spread of twelve centuries.  The late figure comes from Sasanian chronography built on a mythological millennial scheme;  the early ones from the archaism of Gathic Avestan, which is close to Rigvedic Sanskrit.  No Avestan text mentions a datable event.

On the theology, the same article is the single most useful sentence in this whole subject:

Was Zaraθuštra a monotheist, dualist, or polytheist?  These are questions which seem to be as old as the appearance of Zurwanism… and remain the subject of lively discussion today among both scholars and Zoroastrians.  Labels can be misleading, and depending on what one means all three labels can be made to apply to Zaraθuštra's theology.¹³

The Aməša Spənta — Truth, Good Mind, Dominion, Right-mindedness, Wholeness, Immortality — "function as autonomous modalities of Ahura Mazdā's nature."  Call them attributes and you have monotheism;  call them beings and you have a pantheon.  Ārmaiti is called Ahura Mazdā's daughter.¹³

India: one reality, many names

Rig Veda 1.164.46, in Griffith's translation:

They call him Indra, Mitra, Varuṇa, Agni, and he is heavenly nobly-winged Garutmān. To what is One, sages give many a title;  they call it Agni, Yama, Mātariśvan.²⁴

Max Müller coined "henotheism" partly for this.  It is not monotheism:  it does not deny that Indra and Varuṇa exist, it says they are titles of one reality — which is structurally the same move as the Marduk god-list, without the exclusivity.  It is worth listing precisely because it shows how common the move is, and how little the label settles.  Tylor quoted it himself, in the middle of arguing that no savage tribe of monotheists has been known.¹

China: Di, and a live dispute

The Shang oracle bones from Anyang, c. 1200 BC, are among the oldest religious texts anywhere.  They record divinations addressed to royal ancestors and to a power written Di (帝).  David Keightley's standard handbook calls Di "the high god," and lists the divination topics:  whether Di would assist the Shang in war, whether Di would confer harvest, whether Di approved.²⁵

Di has one striking property.  The Shang sacrificed constantly, and lavishly, to their ancestors.  They did not sacrifice to Di.  The high god commands the rain and the harvest and the fate of the city, and receives no cult — the deus otiosus pattern once more, this time in a literate Bronze Age state.

The dispute is over whether Di is a single high god at all, or a title covering a collectivity of ancestral powers.  That question is live in sinology and is not resolved here.

The Sumerian "one tongue," which is not what it looks like

Michael Jones opens his documentary with Enmerkar and the Lord of Aratta, a Sumerian epic containing the so-called Spell of Nudimmud (lines 134–155), which several popular treatments describe as a Sumerian memory of Babel and of a time when everyone worshipped one god.  He flags the problem himself in his video description and calls it "a hook, not part of my main thrust of evidence."⁵  He is right to, and the reasons are worth stating.

The Oxford ETCSL translation reads the passage as a wish for the future, not a report of the past:

At such a time, may the lands of Cubur [Šubur] and Hamazi, the many-tongued, and Sumer… and Akkad… and the Martu land… — the whole universe, the well-guarded people — may they all address Enlil together in a single language!  For at that time… Enki… shall change the speech in their mouths, as many as he had placed there, and so the speech of mankind is truly one.²⁶

Herman Vanstiphout, who edited the poem, states the modern consensus and its limits:  "Many of us now agree that the burden of this spell is that in an ideal world of peace and bliss, which Enmerkar is striving for, only Sumerian will be spoken," while noting in his commentary that "the interpretation of the Spell of Nudimmud is still in dispute.  An alternative reading, which puts it in the past, not in the (immediate) future, is not impossible."²⁷

And even on the older, past-tense reading, the passage says the peoples addressed Enlil with one tongue.  Enlil is one god among many.  The text is about the universality of the Sumerian language and the reach of the Ur III state, not about the number of gods.  It should not be used as evidence of ancient monotheism at all.

The definition problem, which is the real problem

Six of the cases above turn on the same thing.  Is Akhenaten a monotheist, or a henotheist who suppressed rivals?  Is the Marduk god-list monotheism or syncretism?  Is Xenophanes' "greatest among gods" a real plural?  Is Ahura Mazdā's entourage a set of attributes or a pantheon?  Is Rig Veda 1.164.46 monotheism or henotheism?  Is Di one being or many?

None of these is settled by finding more evidence.  They are settled by choosing a definition, and every party to the argument chooses the one that gets the answer they want.  Tylor:  "assigning the distinctive attributes of deity to none save the Almighty Creator" — and therefore no savage monotheists.¹  Schmidt:  created and subordinate beings "do not deserve the title of gods" — and therefore many.¹¹  Same world, opposite counts.

The word itself has a history that explains part of this.  Nathan MacDonald traced it:  "monotheism" was coined by the Cambridge Platonist Henry More in 1660, in An Explanation of the Grand Mystery of Godliness, in the middle of a discussion of pantheism.  Unlike "atheism," "deism" and "polytheism," it is native English rather than a French or Latin import, and it was minted to serve a rationalist theological programme.¹²  MacDonald's argument is that using it as a neutral analytic category for the Old Testament smuggles in "a narrowed, intellectualistic conception of religion" — one that asks how many gods a person believes exist, when the texts are asking whom you will serve.

This is why the Encyclopaedia Iranica's line about Zoroaster is the most honest sentence in the field.  All three labels can be made to apply.  A finding that depends entirely on which one you pick is not a finding about the ancient world.

The practical consequence for this article:  the useful question is not "was culture X monotheistic?" but "does culture X have a supreme creator, is he active, is he moral, and is he alone?"  Those four are separately observable, and the Ethnographic Atlas codes three of them — which is why section 8 could be done at all.

What about Israel?

The standard critical reconstruction runs:  Yahweh began as one of the sons of El in a West Semitic pantheon;  Deuteronomy 32:8–9 in the Qumran and Septuagint text has the Most High apportioning the nations "according to the number of the sons of God," with Yahweh receiving Jacob as his share;  Psalm 82 has God standing in the divine council;  the Kuntillet ʿAjrud inscriptions bless by "Yahweh and his asherah";  and unambiguous denial of other gods' existence appears first in Isaiah 40–55.

That deserves its own article and will get one.  Three things belong here.

First, the Deuteronomy 32:8 reading is not the knockdown it is sometimes presented as, in either direction.  Michael Heiser's treatment — which accepts the "sons of God" text as original against the Masoretic "sons of Israel" — argues that the divine council is a feature of Israelite religion rather than a fossil of an abandoned polytheism, and that Deuteronomy 32 itself subordinates the council to Yahweh.²⁸  Whether that reading holds is a question about Hebrew texts, not about anthropology.

Second, nothing in this article settles it, and nothing in this article is meant to.  The high-god data and the Egyptian and Babylonian texts remove a general argument — "monotheism is late, therefore this text is late."  They do not supply a replacement argument in the other direction.  There is no route from Baiame to Genesis.

Third, and this is the point most worth carrying over:  the same definitional trap operates on Israel.  "Israelite religion was not monotheistic until the exile" and "Israelite religion was monotheistic from Moses" can both be true statements about the same evidence under different definitions of the word — as MacDonald argued at book length.¹²  Anyone arguing that case should say which definition they are using before they start counting.

Objections and responses

"The Ethnographic Atlas is nineteenth-century colonial ethnography with all its biases, so these numbers are worthless."  The biases are real and one of them is documented above:  code 1 conflates absence with non-report, which biases the count against high gods.  But the Atlas is the dataset that every quantitative claim in this field rests on, including the ones used against creationists.  Discarding it discards Swanson, Botero, Watts and Peoples along with everything here.  The right response is to state the biases and their direction, which is what section 8 does.

"Missionary contamination explains the whole ethnographic record."  It does not, and the strongest arguments against it come from Lang and Howitt rather than from anyone modern:  the beliefs were recorded ahead of missionary presence in specific documented cases, they were held as initiation secrets from precisely the outsiders who would have taught them, and the tribes with the longest mission contact are the ones without the belief.³⁶  What contamination does explain is a great deal of the modern pattern, which is why 63% of the Atlas's morally active high gods sit in the Islamic and Christian belt.

"But nine of ten Australian societies in the Atlas have no high god, and Australia was Lang's best case."  True and important — and the ten do not include a single south-eastern group, which is where the case was made.  See section 8.  This cuts against confident claims in both directions.

"Akhenaten does not count;  it was one king, it lasted seventeen years, and it was erased."  All true, and all of it is in section 11 above.  The claim being made is narrow:  a documented, exclusive, one-god religion existed in the fourteenth century BC.  That is enough to defeat "monotheism could not exist that early," which is the claim it is being used against.

"The Marduk god-list is syncretism, not monotheism."  Finkelstein said exactly that, and Lambert quotes him saying it before disagreeing.⁹  The dispute is real and unresolved.  It is also a dispute about the word, not about the tablet:  everyone agrees the tablet makes the major gods into aspects of one god.

"None of this proves God revealed himself to early man."  Correct, and this article does not argue that it does.  Michael Jones says the same thing in his own video, at length, and disagrees with Schmidt about it explicitly.⁵  Howitt, whose data the argument rests on, ruled it out in 1904.³  Lang ruled it out too:  his account of how the idea arose is natural theology — the inference from made things to a maker — not revelation.²

"Then what is the argument even for?"  For deleting a premise.  A great deal of critical scholarship, and nearly all popular skeptical argument, still assumes an evolutionary sequence in which monotheism is a late achievement.  That premise is doing real work in dating texts and in evaluating what a Bronze Age person could plausibly have believed.  It is not supported.  Removing it does not prove anything about Genesis;  it means the question has to be argued on the texts.

Verdict

Against the apologetic claim.  Original monotheism is not established and cannot be, and should be dropped as an argument.  Foragers are the least likely people in the ethnographic record to hold a high god — 39.5% against 64.3% overall, and 5.9% against 25.5% for an active moral one.  The gradient runs with political complexity, not against it, at rho = 0.44.  Schmidt's method is dead, his ranking of "primitive" cultures was revisable after the fact, and the best modern reconstruction puts animism at the root.  Howitt, the argument's own principal source, denied the theological reading in print.

Against the skeptical claim.  The ladder is gone, and its most-cited modern support had to be retracted.  Belief in a supreme creator does not correlate with any other religious trait — not ancestor worship, not even belief in an afterlife — which is fatal to the specific mechanism Tylor and Spencer proposed.  It appears in about two of every five foraging societies on every inhabited continent.  Where foragers have a creator, three times in four he is the withdrawn creator, a type Tylor's three mechanisms do not naturally produce.  And in the literate world, exclusive monotheism is documented in Egypt by about 1350 BC, argued philosophically in Greek by about 530 BC, and asserted of Marduk in Babylonian by at least 800 BC.  Nothing about a monotheistic text requires it to be post-exilic.

On the field itself.  The most-quoted numbers in this literature rest on one variable whose "absent" category means "absent or not reported," on a sample that excludes the societies the argument was originally about, and on a moralizing-god count that is 63% Islam and Christianity.  The retraction of the 2019 Nature paper over a missing-data decision is not an anomaly;  it is what this evidence base is like.  Every figure in this article, including the ones computed here, should be held that loosely.

What survives, and it is worth having.  Two claims:

1. The withdrawn creator is not a product of civilization.  He is commonest, proportionally, among the simplest peoples and in the oldest literate pantheons — An at Nippur, Di at Anyang, Baiame in the sky-country.  What civilization adds is not the creator.  It is the enforcement.

2. Monotheism has no fixed place in a sequence, because there is no sequence.  It shows up in the fourteenth century BC and in the sixth, in an Egyptian palace and in a Greek poet and in a Babylonian god-list, and it disappears again.  Whether Israel's belongs early or late is a question about Israelite evidence, and it has to be answered there.

Method and data

The recount uses the D-PLACE CLDF release of the Ethnographic Atlas (Murdock et al. 1999), 1,291 societies, CC-BY-NC 4.0.²⁹  The analysis script is research/high-gods.ts beside this article — Deno TypeScript, no dependencies, reads the four CLDF CSVs and prints every table above plus the full by-region breakdown.  Fetch the data and run it:

mkdir -p dplace && cd dplace
for f in data.csv societies.csv variables.csv codes.csv; do
  curl -sLO "https://raw.githubusercontent.com/D-PLACE/\
dplace-dataset-ea/main/cldf/$f"
done
cd .. && deno run --allow-read research/high-gods.ts dplace

Variables used:  EA034 high gods, EA042 dominant subsistence, EA030 settlement pattern, EA031 mean community size, EA033 jurisdictional hierarchy beyond the local community, EA066 class differentiation.  Correlations are Spearman's rho on the ordinal codes;  the reported p-values assume independent observations and are therefore too generous, since neighbouring societies share ancestry (Galton's problem).  The class-differentiation ordinal is not cleanly monotonic — Murdock's "elite" and "dual" stratification are not ordered against each other — so its rho should be read as indicative only.

The thirteen regions counted as the Islamic and Christian belt are named in the script and in section 8;  the choice is a judgment call, and the count is reported both with and without the Indian Subcontinent.

Sources